A new fluorescence in situ hybridization (FISH)-based method for multiplex RNA imaging is reported in Nature Methods this week. Developed by scientists at Singapore's Agency for Science, Technology, and Research, the approach uses split probes wherein a bridge sequence is shared between a pair of adjoining encoding probes. "Only when a pair of encoding probes are hybridized at adjacent locations on the target RNA will there be sufficient complementary base pairing in close proximity to enable the bridge probe to bind efficiently," the researchers write. "A fluorescently labeled readout probe then hybridizes to the bridge probes to generate on-target signals." After optimizing the probes and workflow, the team demonstrates their technique — called split-FISH — in various mouse tissues to quantify the distribution and abundance of 317 genes in single cells and uncover diverse localization patterns for spatial regulation of the transcriptome in complex tissues.
Members of the Genome in a Bottle Consortium report in Nature Biotechnology this week a benchmark for the detection of false-negative and false-positive germline large insertions and deletions. The group developed benchmark structural variant calls and benchmark regions for a son in a broadly consented and available Ashkenazi Jewish trio from the Personal Genome Project by integrating 19 sequence-resolved variant calling methods from diverse technologies. The scientists' final benchmark set contains 12,745 isolated, sequence-resolved insertion and deletion calls greater than or equal to 50 base pairs, enabling anyone to assess both false-negative and false-positive rates, they write. GenomeWeb has more on this here.
A novel network-based gene set embedding approach for gene set analysis is presented by a trio of Stanford University investigators in Nature Machine Intelligence this week. Called Set2Gaussian, the method takes biological networks and a collection of gene sets as input, then finds a low-dimensional space in which genes and gene sets preserve their distances in the network, the researchers say. Each gene is represented as a single point in the low-dimensional space, and each gene set is represented as a multivariate Gaussian distribution that is parameterized by a mean vector and covariance matrix. In the researchers' experiments, Set2Gaussian improves gene set member identification, accurately stratifies tumors, and finds concise gene sets for gene set enrichment analysis. The team also uses it to identify a clinical prognostic and predictive subnetwork around neurofilament medium in sarcoma, which they validate in independent cohorts.